This repository hosts the trained models in Handlan and Sheng (2023) as well as a Jupyter notebook demo of how to apply the trained models to classifying audio clips.
The demo notebook can be found under the demo folder. The demo uses two audio clips from the NBER 2023 Summer Institute Methods Lectures. These lectures are publicly available and downloadable on Youtube. The downloaded audio clips are stored in data/NBER.
The trained models are under the model folder. The models are trained with 5-fold cross validation for each gender and label. There are a total of 25 models.
If you want to use our models, we would appreciate you citing our paper as the following:
Handlan, Amy and Sheng, Haoyu, Gender and Tone in Recorded Economics Presentations: Audio Analysis with Machine Learning (January 1, 2023). Available at SSRN: https://ssrn.com/abstract=4316513 or http://dx.doi.org/10.2139/ssrn.4316513